Statistical Methods for Biomarker Performance Evaluation

• Activity Number: 542
• Type: Invited
• Date/Time: Wednesday, August 1, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): David Raunig*+
• Companies: ICON Medical Imaging
• Keywords: biomarker qualification ; performance ; ROC Curve ; AUC ; discriimination index ; longitudinal analyiss

Abstract

Biomarkers must successfully demonstrate that they can consistently arrive at the correct conclusion and that effort is necessarily separated into evaluations of technical and clinical performances. Technical performance evaluates consistency, is concentrated on a continuous measurement evaluation of bias, precision and linearity and is the foundation for successful Clinical performance. Clinical performance evaluates the ability of the biomarker to successfully predict clinical outcome and is traditionally evaluated over the entire dynamic range of the biomarker with the Area Under the ROC Curve (AUC). However, AUC is only part of the story and neglect critical components of the final evaluation of the biomarker for use in a clinical trial. Novel methods consider the sensitivity and specificity together but do not regard that some regions of the ROC are more critical than others in a clinical trial. We will discuss the results from the biomarker regulatory qualification of novel hepatotoxicity biomarkers for a longitudinal model for necropsy and to demonstrate the utility of a risk-weighted vector-based assessment of clinical performance in a preclinical model